Indicators for the Number of Females Choosing STEM Majors

John Carroll University John Carroll University

Carroll Collected Carroll Collected

Senior Honors Projects Honor's Program

Spring 2014

Indicators for the Number of Females Choosing STEM Majors Indicators for the Number of Females Choosing STEM Majors

Olga Graves

John Carroll University

, ograv[email protected]

Follow this and additional works at: https://collected.jcu.edu/honorspapers

Part of the Education Commons

Recommended Citation Recommended Citation

Graves, Olga, "Indicators for the Number of Females Choosing STEM Majors" (2014).

Senior Honors

Projects

. 34.

https://collected.jcu.edu/honorspapers/34

This Honors Paper/Project is brought to you for free and open access by the Honor's Program at Carroll Collected.

It has been accepted for inclusion in Senior Honors Projects by an authorized administrator of Carroll Collected. For

more information, please contact mchercour[email protected].

JOHN CARROLL UNIVERSITY

Indicators for the Number of Females

Choosing STEM Majors

Olga Graves, Undergraduate; Dr. Andrew Welki,

Economics & Finance (Advisor)

Economics Capstone Paper

Senior Honors Project

EC 410 Paper

4/24/2014

Abstract

This paper explores the different variables which may motivate females to choose STEM

(science, technology, engineering, and math) majors. Historically, women have been greatly

underrepresented in STEM fields for a number of different cultural and economic reasons.

However, in order to fully compete in the global economy, the United States must find a way to

bolster female participation in these fields. The motivators chosen to explore in this paper are:

female faculty numbers, federal financial obligations, early concentration in math, SAT math

scores, appropriations through the Women’s Educational Equity Act, average salary for STEM

occupations, and female unemployment rates. The study found that all variables except federal

financial obligations and the female unemployment rate had the expected sign and were

statistically significant. The paper proposes that in order to create a more complete model for

predicting the number of female STEM majors, cultural trends and attitudes should be an

considered.

Introduction:

On June 23, 1972 Title IX was signed into law by President Richard Nixon. This was

one of the first federal laws which acknowledged the fact that women were underrepresented in

education, but helped to specifically highlight the gender gap in science, technology,

engineering, and math (STEM). This underrepresentation of women in STEM fields, though it

has improved, has persisted throughout the years and still exists today. Women constitute a

dramatically smaller portion of certain STEM-heavy fields than men. In 2009 the Department of

Commerce found that, though women constitute close to half of the overall workforce, they hold

less than 25 percent of STEM jobs (Beede 1). And in 2011 the Bureau of Labor Statistics cited

that women make up 81.7 percent of elementary and middle school teachers, but only 33.9

percent of computer systems analysts and, even more astonishingly, only 4.3 percent of flight

engineers.

With an increasing dependence on technology in the workforce, female participation in

STEM fields is an essential step towards leadership equality between genders in the workforce.

Gaining more women in STEM fields would also increase innovation and accelerate the United

States’ ability to compete globally in many markets.

The fact that fewer women are receiving bachelor degrees in STEM fields has been cited

as a good indicator as to why there are fewer women in STEM occupations because one of the

strongest indicators for occupation selection is a person’s undergraduate degree (Griffith 1). A

study was done by the National Academy of Sciences which found that “education at the

undergraduate level is vital to developing a workforce that will allow the United States to remain

the leader in the 21

century global economy.” This means that in order to remain competitive

and productive, we must gain more participation in STEM majors so as to bolster participation in

STEM occupations (S. 3475).

In order to explore this dilemma of motivating women to join STEM majors, this paper

will look at the different indicators which influence an individual to choose a specific major.

The general indicators, chosen from previous academic research on the subject, will be applied to

the problem of how to more specifically motivate females to choose STEM majors. It may be

helpful to better understand these specific variables so that steps can be taken to further bolster

female participation in STEM occupations later in life.

Literature Review:

One factor that is supported by research to have an effect on choice of major by women

and which may encourage women to attain degrees in STEM fields is the prevalence of female

faculty in those fields. In a study on the effect that having female professors whom students see

as role models has on women pursuing science majors, Young and her associates found that

“women with a female professor showed a stronger implicit science identity to the extent they

viewed her as a positive role model” (Young 288). Similar research has been done in other fields

such as mathematics which points to the fact that women and girls who are taught by female

instructors identify more with mathematics, earn higher grades in mathematics, and thus have

more confidence in the field (Stout 260). Therefore, it will be crucial to explore the change in

the number of female faculty members across time in order to determine if an increase in female

faculty members in STEM departments had an effect on female STEM graduates.

One variable that will be explored, but which did not have substantial academic literature

was the degree to which federal financial obligations to universities for STEM education

determines recruitment or retention of women in STEM fields. Because the vast majority of

federal initiatives being taken to induce women to join STEM fields are monetarily based in

grants, the effects of such grants will be interesting to pursue.

Another factor which greatly affects women’s participation in STEM programs during

their postsecondary education is the level of interest which they have in those fields when

entering college. A person is much more likely to explore a major and furthermore a career in a

subject in which they have experience. This theory proves to be especially true for females in

quantitative subjects. In their study on expressed interest in STEM fields upon graduation of

high school, Amy Bergerson and her colleagues found that early intervention was key to creating

interest in quantitative fields in high school females, most specifically in engineering (Bergerson

611). Therefore, studying the effect of concentration on and performance in mathematics in

elementary school will be crucial to determining if early intervention helps create more interest

in females in STEM degrees.

Another factor which was explored in this same study was the idea that “strong past

achievement is likely to be associated with strong positive self-efficacy beliefs, which, according

to the literature, are potent determinants of behavioral initiation and persistence” (Bergerson

611). This means that students who do well in a specific subject tend to choose to further their

participation in that subject. Another study which echoes this idea that students become more

interested in subjects in which they excel was a study done by Lindsay Calkins and Andrew

Welki in which they explored the factors which help students decide to major in economics. In

their study they found that “positive reinforcement” supplied through the achievement of good

grades is a strong motivator for persisting in a major (Calkins 6). By comparing the trend of

SAT scores for females on the math portion of the exam, this paper will attempt to determine the

effect that higher scores in STEM subjects incentivize women to major in STEM fields while in

college.

In a study during which researchers surveyed students to find their expected earnings

after graduation with a particular major, economists found that expected earnings do have an

effect on college major choice (Arcidiacono 25). They found that students were more likely to

choose a major in which the average student is expected to earn more than in other majors

(Arcidiacono 25). Additionally, in separate research, economists from Georgetown University

found that “majors with high technical, business and healthcare content tend to earn the most

among both recent and experienced college graduates” (Carnevale 6). Both of these studies point

to the idea that because STEM majors tend to earn higher salaries than other majors, more

students may choose STEM degrees. Therefore, the variable of average salary for STEM

occupations will be utilized in this paper.

As stated earlier, Title IX exists to protect a person from being discriminated against

because of their gender in education programs which receive federal financial assistance (“Title

IX and Sex Discrimination”). After Title IX was enacted in 1972, the Women’s Educational

Equity Act of 1974 was initiated in order to more definitively specify how Title IX would

promote gender equity in education (Heston-Demirel). The WEEA was charged with providing

financial assistance to educational agencies and programs to help meet certain gender equity

requirements laid out in Title IX. Under continued reauthorizations, the WEEA continues to

provide funds for equity programs as well as support technical assistance to implement those

programs (“Subpart 21 - Women’s Educational Equity Act”). Because this is the most

substantial piece of federal legislation which exists to promote and protect equality in education

for all genders, it will be crucial to include its existence for review in the model.

It has been found that nontechnical majors have a higher unemployment rate than

technical majors. For example, in a study done in 2012 on the different unemployment rates

received by different majors, Anthony Carnevale and his colleagues found that “majors… related

to technical occupations tend to have lower unemployment rates than more general majors, like

Humanities and Liberal Arts, where graduates are broadly dispersed across occupations and

industries” (Carnevale 5). They found that because technical degrees such as healthcare and

physical sciences are so specifically tied to certain occupations, they tend to have lower

unemployment rates. Furthermore, in his research on income and degree choices, researcher

Jacopo Mazza found that degrees in the sciences “offer better job security in times of economic

uncertainty” (Mazza). Logically, because science degrees offer more stability, it may be

assumed that students are more likely to choose STEM majors for an added sense of job security

when the unemployment rate is higher. Students wishing to have a stable job after graduation

may be more inclined to choose a more technical major in order to increase their chances of

getting a job. This feeling may be exacerbated during times of high unemployment rates.

Therefore, data for unemployment rates over time will be explored in this paper.

After reviewing the literature on possible factors that can be manipulated in order to

incentivize women to major in STEM fields and thus, work in STEM fields, some possible

indicators to explore have been narrowed down. These determinants are: the number of female

faculty members in STEM fields, the size of federal financial obligations dedicated to university

STEM programs, early concentration in mathematics, math SAT scores, average expected salary,

appropriations through the WEEA, and unemployment rates.

Model:

The functional form for the linear regression model for determining female STEM majors

as a percentage of total female bachelor degree recipients is postulated as follows:

FemSTEM = ƒ (FemProf, FedFin, NAEP, SATMath, AvgSal, WEEA, FemUnemp)

Table 1: Variable Definitions and Sources

Variable Definition Data Collected From:

FemSTEM Female STEM majors as a percentage of

total female BA recipients

National Center for Science

and Engineering Statistics

FemProf Number of female professors teaching in

STEM fields

National Science Foundation

FedFin Federal financial obligations designated

for promoting STEM majors

National Science Foundation

NAEP Math score on National Assessment of

Educational Progress exam for fourth

grade female students

National Center for Education

Statistics

SATMath Average scores for females on SAT math

exam

The College Board

AvgSal Average expected salary for STEM field

majors

National Science Foundation

WEEA Enactment of the Women’s Educational

Equity Act

U.S. Department of Education

FemUnemp National unemployment rate 2013 Economic Report to the

President

The table below contains the expected affect that each independent variable will have on

the dependent variable. Each expected sign also comes with a brief explanation as to why the

expected sign was chosen.

Table 2: Expected Variable Effects

Variable Expected Sign

FemProf +

FedFin +

NAEP +

SATMath +

AvgSal +

WEEA +

FemUnemp +

The predicted effect of FemProf on FemSTEM is direct. As stated previously, the theory

suggests that women are more likely to choose a major when they are able to visualize

themselves in that role. Having female faculty members in STEM fields will increase the

number of role models for female students, thus increasing the number of female students

choosing to major in STEM fields.

The predicted effect of FedFin on FemSTEM is direct. A higher federal financial

obligation to STEM programs in universities will provide students with better facilities and

training programs and thus a better overall experience in their STEM classes. Included in federal

financial obligations is support for research and development facilities, facilities for instruction,

fellowships and training grants, among other things. These programs are put in place to bolster

the experience of students and encourage further participation in STEM program. I predict that

more investment will likely lead more female students to choose to major in STEM.

The predicted effect of NAEP on FemSTEM is direct. As the theory shows, early

concentration in a subject is key to sustaining and cultivating that interest further later in life.

Early intervention will peak students’ interest and build the platform for further achievement in

STEM subjects. Because achievement in STEM subjects tends to lead to majoring in the field,

more concentration earlier on will lead to more STEM majors. Therefore, I predict that a higher

score on the math portion of the NAEP test given to female fourth graders will lead to more

female STEM majors.

The predicted effect of SATMath on FemSTEM is direct. Achievement in a subject often

leads to concentration in that subject, as supported by theory. Therefore, higher math SAT scores

for women will indicate a higher achievement and preparedness in the subject, suggesting an

increase in STEM majors.

The predicted effect of AvgSal on FemSTEM is direct. One reason for choosing a major,

as found during review of the literature, is the potential earning power of that major. STEM

majors tend to earn a higher salary than other majors, such as the humanities. A high expected

salary may lead more women to major in STEM fields.

The predicted effect of WEEA on FemSTEM is direct. Providing federal support to the

issue of women in education will increase funding to programs that push women to join

nontraditional educational programs.

The predicted effect of FemUnemp on FemSTEM is direct. It has been proven that the

unemployment rate of recent college graduates differs depending on a student’s major choice.

Technical majors such as those found in STEM have a lower unemployment rate upon

graduation than other majors such as arts and humanities. The national unemployment rate is

often called upon as a figure for assessing the health of the nation’s economy and workforce. By

looking more specifically at the female unemployment rate, we may see an effect on the number

of females joining STEM majors. When the female unemployment rate is high, I hypothesize

that students will allow the desire to be more hirable affect their decision on what to major in. It

is for this reason that I believe that a higher female unemployment rate will cause more women

to turn to STEM majors for job security.

Data:

Female STEM Majors

The dependent variable (FemSTEM), Female STEM majors as a percentage of total

female BA recipients, has been calculated from a data set provided by the National Center for

Science and Engineering Statistics. This data is available from 1966 to 2010. Because the

number of women receiving bachelor’s degrees has increased over time due to several different

factors, turning the actual number of female STEM majors into a percentage of the total number

of female degree recipients will help to correct for this general upward trend. In order to

determine that the proposed factors are working, I will thus be looking to see that the proportion

of women who are receiving bachelor’s degrees in STEM fields is increasing in order to

determine an upward trend.

Female Faculty

The number of female faculty members teaching in STEM fields (FemProf) was gathered

from the National Science Foundation and indicates the number of female science, engineering,

and health doctorates who are employed in academia. This data is accountable for women who

have earned their doctorate degrees from an American university only. It is available for every

other year from 1973 to 2010. It is stated in thousands. In order to allow for the inclusion of this

data in the full regression model, the years for which the number of female faculty in STEM

fields is missing have been calculate manually as the mean of the years before and after the year

in question. For example, the data cited for 1974 is calculated as the mean of the values for 1973

and 1975. This data will be used to determine the degree to which having female role models

affects the number of female STEM majors.

Federal Obligations

Federal financial obligations (FedFin) is represented in billions of 2014 dollars by the

total federal obligations for science and engineering to universities and colleges. This data was

collected from the National Science Foundation and is available from 1963 to 2011. It includes

all the aggregated federal obligations, though the data can be split up by activities such as:

research and development, R&D plant, facilities for instruction in science and engineering,

fellowships, traineeships, and training grants, general support for science and engineering, and

other science and engineering activities. For the purpose of this study, I will be using the

aggregate of all of these activities.

Average National Assessment of Educational Progress

The variable EarlyMath will be represented by the mathematics scores for female fourth

graders on the National Assessment of Educational Progress. The data was collected from the

National Center for Education Statistics. It is available sporadically beginning in 1973 through

2012 and is measured in an averaged test score from 0 to 500. Due to the irregularity of the data,

these values will be regressed on their own against the dependent variable rather than including

them in the full regression. Though the data is not included in the full regression model, due to

theoretical support in the literature review process, it will remain in this paper because it is

believed to have a significant effect on female STEM majors.

Furthermore, though the data for the NAEP scores is reported beginning in 1973, each

year’s score will be matched with the dependent variable value of twelve years later. For

example, the 1973 test scores for fourth grade females will be match with the output of 1985

female STEM majors. This is due to the fact that the girls whose tests scores are reported each

year, in this example 1973, will be more appropriately matched with the outcome of females

graduating college and determining their bachelor degree field, in this example during the year

1985. This will be represented in the data table presented for the regression (Table 6). The

NAEP data will be used to test the “early achievement” hypothesis that early focus and

achievement in a subject encourages students to major in them.

SAT Math Scores

The variable SATMath is represented by the average yearly score for women on the

mathematics portion of the SAT exam. This data was collected from The College Board and is

available from 1967 to 2013. Initially this study was going to use data from the ACT test in the

subjects of science and math. Data was collected on both portions of the ACT exam, however a

test of correlation between ACT math and science test scores revealed a correlation of 0.848

between the two meaning that they were highly correlated. Because skills in mathematics and

quantitative knowledge are often required in order to excel in science fields, this paper will omit

the ACT science scores. I have chosen to use SAT math scores instead of ACT math scores as

an indicator of strong past achievement in STEM subjects because SAT scores available for a

larger number of years.

Excepted Salary

The average salary per year for STEM majors, denoted at AvgSal, was collected from a

number of reports cited in the Science and Engineering Indicators report put out by the National

Science Foundation. The data is reported in 2014 dollars and is calculated for some years based

on growth projections of previous years. It is not broken down by gender and thus contains the

average salary for both male and female STEM workers.

Women’s Educational Equity Act

The variable WEEA is a representation of The Women’s Educational Equity Act which

began appropriating funds to support women’s equality in educational fields in which they are

underrepresented in 1976. Though Title IX was enacted in 1972 and the Women’s Educational

Equity Act was first enacted in 1974, actual funding through this act to educational programs did

not begin until the fiscal year of 1976 (U.S. Department of Education). Through reauthorization,

this Act has been continually funded since its inception except for during the year 1996. The

appropriation amount is used for data here and is listed in millions of dollars. From this variable

we will better understand if and to what degree the funding from the WEEA helped motivate

women to join STEM majors.

National Unemployment Rate for Females

The national unemployment rate for females, denoted as FemUnemp, was collected from

the 2013 Economic Report of the President. The statistics were available from 1966-2012 and

were reported as a percentage of civilian labor force in the group specified (in this case the group

was females).

The following table contains selected variable statistics:

Table 3: Summary Statistics on Variables

Variable Mean Standard Deviation

FemSTEM

(percent)

26.150 1.597

FemProf

(thousands)

49.0 27.741

FedFin

(billions of

2014 dollars)

21.792 8.174

NAEP 226 5

SATMath 487.263 10.246

AvgSal 82,397 2,806.95

WEEA

(millions of

2014 dollars

8.250 8.972

FemUnemp 6.4 1.473

Estimation Results:

Three regressions were run using the data collected for each variable referenced

throughout this paper. The reasoning behind running three regressions was because of the

unavailability of data for the NAEP test scores as well as limited availability of average salary

data for STEM workers.

The NAEP test scores are only available for sporadic periods of time. Therefore, it was

difficult to run a regression using the NAEP test scores in a model alongside all of the other

variables without losing much of the data for those other variables.

Additionally, salary data was only available for years beginning in 1993 onward.

Because of this, it was decided to run a regression which did not include the salary data, but did

include all the available data for the rest of the variables which stretched back to 1973. When

assessing the model which included the salary data, it was found that the variable AvgSal was

statistically insignificant with a p-value of .346. Additionally, many of the variables such as

FemProf, FedFin, WEEA, FemUnemp became statistically insignificant when regressed in a

model with AvgSal. It is for these reasons that the model run with the salary data will not be

discussed further in this paper

and the variable of average salary will not be included in the

Regression One.

The data table including average salary data is available in Appendix A. The regression results for the regression

run including salary data are included in Appendix C.

Regression One: FemProf, FedFin, SATMath, WEEA, FemUnemp

Table 4: Regression One Data

Year FemSTEM FemProf

FedFin

SATMath WEEA

FemUnemp

1973 23.486% 10.7 $13.031

489 $0.000 6.0

1974 24.223% 12.2 $13.034

488 $0.000 6.7

1975 24.292% 13.6 $12.244

479 $0.000 9.3

1976 24.401% 15.1 $12.212

475 $25.871 8.6

1977 24.468% 16.5 $12.983

474 $28.165 8.2

1978 24.518% 18.0 $14.268

474 $29.113 7.2

1979 24.428% 19.4 $14.446

473 $29.105 6.8

1980 24.536% 21.3 $13.650

473 $28.492 7.4

1981 24.509% 23.1 $13.076

473 $20.985 7.9

1982 24.954% 24.8 $12.601

473 $14.013 9.4

1983 24.773% 26.5 $13.390

474 $13.577 9.2

1984 25.047% 28.8 $14.254

478 $13.015 7.6

1985 25.559% 31.1 $15.836

480 $13.091 7.4

1986 25.617% 32.6 $15.919

479 $12.296 7.1

1987 25.369% 34.0 $17.708

481 $7.233 6.2

1988 24.949% 36.4 $18.135

483 $6.650 5.6

1989 24.600% 38.7 $19.086

482 $5.583 5.4

1990 24.725% 40.3 $18.809

483 $3.768 5.5

1991 24.764% 41.9 $20.526

482 $3.438 6.4

1992 25.531% 44.4 $21.483

484 $0.836 7.0

1993 25.823% 46.9 $20.852

484 $3.223 6.6

1994 26.464% 49.7 $21.968

487 $3.143 6.0

1995 27.349% 52.4 $22.277

490 $3.178 5.6

1996 27.820% 55.8 $21.622

492 $0.000 5.4

1997 28.279% 59.2 $22.082

494 $4.388 5.0

1998 28.254% 61.8 $23.179

496 $4.321 4.6

1999^ 28.110% 64.4 $25.446

495 $4.227 4.3

2000 27.966% 67.5 $27.101

498 $4.090 4.1

2001 28.035% 70.5 $29.833

498 $3.979 4.7

2002 28.050% 74.6 $31.883

500 $3.915 5.6

2003 28.337% 78.7 $34.017

503 $3.802 5.7

2004 28.290% 84.4 $33.978

501 $3.681 5.4

2005 28.209% 87.3 $33.710

504 $3.553 5.1

2006 27.798% 90.1 $32.916

502 $3.408 4.6

2007 27.538% 92.2 $31.911

499 $2.128 4.5

2008 27.514% 94.2 $30.996

499 $2.013 5.4

2009 27.457% 99.7 $39.578

498 $2.615 8.1

2010 27.677% 105.2 $38.037

499 $2.608 8.6

Mean 26.150% 49.0 $21.792

487.263 $8.250 6.4

StdDev 1.5967% 27.7414 $8.174 10.246 $8.972 1.473

in thousands, every other year starting with 1974 is an average

in billions, converted to 2014 dollars

in millions, converted to 2014 dollars

The functional model for this linear regression is stated as follows:

FemSTEM = ƒ (FemProf, FedFin, SATMath, WEEA, FemUnemp)

Table 5: Regression One Estimates (Dependent Variable: FemSTEM)

Independent

Variable

Expected

Sign

Estimated

Coefficient t-statistic P-value

Intercept -0.276

-1.559

0.129

FemProf

0.000658

3.536

0.00126

FedFin

-0.00160

-2.232

0.0327

SATMath + 0.00111

3.025

0.00487

WEEA + 0.000369

2.099

0.0438

FemUnemp + -0.000574

-0.541

0.592

.864

Adjusted R

.843

Based on the R

value of this model of .864, this regression is a good fit. Generally

speaking, an R

value of above a .7 can be considered a “good fit.” The R

value of a regression

represents the proportion of the variation in the dependent Y variable explained by the set of

independent X variables. An R

of .864 suggests that 86.4% of the percentage of females in

STEM majors is explained by the set of independent variables used in the model (number of

female professors, federal financial obligations, SAT math scores, WEEA appropriations, and

female unemployment rate).

In order to test further the usefulness of this model, a test was run using the calculated R

value of .864 to calculate the overall F statistic of 40.659

. The critical F value for a model with

a numerator of 5 degrees of freedom and denominator of 32 degrees of freedom is F

(5,32)

: 2.05 at

a significance level of 10 percent. The calculated F value for this model is larger than the critical

F value, meaning that it has statistical significance and the independent variables do have an

effect the dependent variable.

Equation used to find overall F value: (R

/k-1)/((1-R

)/(n-k))

In addition to the R

value, the p-values can be used to show the strength of the individual

independent variables on the dependent variable. A p-value that is very small (generally, less

than .05) usually indicates a stronger relationship with the dependent variable. In this regression,

the variables FemProf, FedFin, SATMath, and WEEA have very small p-values and thus can be

seen as having a strong relationship with the Y variable of FemSTEM. The intercept value and

the variable FemUnemp, however, have larger p-values (.129 and .592 respectively) and

therefore can be seen as having weaker relationships with the dependent Y variable.

The estimated coefficient for FemProf is approximately 0.000658 resulting in a positive

relationship between the number of female faculty members in STEM fields and the percentage

of women choosing STEM majors. This coefficient means that an increase in 1,000 female

professors will lead to a 0.000658 percent increase in female STEM majors. This result is

consistent with the model prediction. It supports the theory that more women in faculty roles in

STEM fields allows more female students to see women in these skilled positions, allowing

female students to more easily picture themselves in those roles and causing them to stick with,

or choose, STEM majors more frequently.

The estimated coefficient for FedFin is approximately -0.00160 resulting in a negative

relationship between the amount of federal financial obligations designated for promoting STEM

fields in universities and the percentage of women choosing STEM majors. This result is not

consistent with the model prediction. It was predicted that there would be a positive relationship

between the amount of money provided by the federal government and the number of female

STEM majors. This finding of a negative relationship between FedFin and FemSTEM could be

caused by a number of things. First, the data for FedFin may be regressed against the wrong

values and years for FemSTEM. It may be possible that the money dispensed each year towards

STEM fields does not take effect on female major choices until a year or two later. Therefore, it

may be more statistically correct to attribute the amount of money dispensed one year, on the

FemSTEM percentage outcome of the following year

. A second item that may be throwing off

the sign for FedFin is the large jump in funding from $30.996 in 2008 to $39.578 in 2009. This

is due to the fact that the data used for the regression includes funding dispersed in 2009 that was

designated under the American Recovery and Reinvestment Act of 2009. This stimulus

package put forth a total of $1 billion towards education alone, some of which was further

designated specifically for STEM fields, causing the large jump from 2008 to 2009. This is a

small bump in the trend, however, and most likely does not account for the full error in

prediction. Though it does not have the expected sign, because it is statistically significant, it

will remain in the model.

The estimated coefficient for SATMath is approximately 0.00111, resulting in a positive

relationship between the average math SAT scores for females and the percentage of women

choosing STEM majors. This means that a one point increase in the average math score for

females will lead to a .00111 increase in the percent of female STEM majors. This supports the

theory that achievement in a subject will encourage students to focus on that subject. Thus, a

higher score on the math portion of the SATs can lead to a higher percentage of females

choosing STEM majors.

The estimated coefficient for WEEA is approximately 0.000369, resulting in a positive

relationship between the appropriations set for the Women’s Educational Equity Act and the

percentage of women choosing STEM majors. This means that a $1 million increase in WEEA

appropriations will lead to a 0.000369 increase in the percent of female STEM majors. This

For example, the 2008 value for FedFin of $30.996 may be more useful if it were regressed against the 2009 value

for FemSTEM of 27.457%.

coefficient supports the model predictions and the theory that higher appropriations that promote

women specifically in education will lead to more women in STEM fields.

The estimated coefficient for FemUnemp is approximately -0.000574, resulting in a

negative relationship between the female unemployment rate and the percentage of women

choosing STEM majors. This does not correspond with the expected sign. It should also be

noted, however, that the p-value of .592 renders the variable statistically insignificant in this

model. One reason that the predicted sign may be wrong and the variable may be statistically

insignificant could be that the reported annual unemployment rates may not affect actual

behavior until years later. When a person is laid-off it may take some time for them to go back

to school, potentially holding previous years of unemployment responsible for current years

STEM rates

. Because theory strongly suggests that in times of high unemployment people tend

to choose STEM majors for their stability and strong job opportunities, this variable will remain

in the model.

Regression Two: NAEP Scores

Table 6: Regression Two Data

Year for

STEM

Majors

FemSTEM

Year Test

Taken

NAEP

1985 25.559% 1973 220

1990 24.725% 1978 220

1994 26.464% 1982 221

1998 28.254% 1986 222

2002 28.050% 1990 230

2004 28.290% 1992 228

2006 27.798% 1994 230

2008 27.514% 1996 229

This concept is similar to that discussed with FedFin. Some of the females choosing STEM majors in 2010 may be

attributable to female unemployment rates from 2006, for example.

The functional model for this linear regression is stated as follows:

FemSTEM = ƒ (NAEP)

Table 7: Regression Two Estimates (Dependent Variable: FemSTEM)

Independent

Variable

Expected

Sign

Estimated

Coefficient t-statistic P-value

Intercept -0.211

-1.164

0.289

NAEP

0.00214

2.658

0.0376

.541

Adjusted R

.464

As stated earlier, due to the fact that the NAEP test has been administered sporadically

for the past thirty years, it was necessary to regress, or correlate, the NAEP math scores alone for

each of the available years to the percentage of females in STEM majors. The NAEP math

scores listed here are the average for females in fourth grade. In Table 6, it can be seen that the

values for FemSTEM in each year are matched with the NAEP math scores for twelve years

prior. This is so that the female students who took the NAEP in fourth grade would be the

traditional age of graduating seniors in college. By regressing the numbers this way the NAEP

scores are more closely attributed to the class that took them.

Though the R

value is .541, this is a suitable number for a regression with only one

variable. The p-value for the NAEP scores is .0376, making it statistically significant.

The estimated coefficient for NAEP is approximately 0.00214, resulting in a positive

relationship between the scores achieved by female students on the math portion of the NAEP

and the percentage of females choosing STEM majors twelve years later. This finding supports

the model prediction that early achievement in math, recorded by higher test scores in the math

NAEP by females, will lead to more females choosing to focus on STEM subjects in college.

Standardized Coefficient Model

Table 8: Standardized Coefficient Estimates

Independent

Variable

Standardized

Coefficient

FemProf

1.144

FedFin

-.821

SATMath

.712

WEEA

.208

FemUnemp

-.0529

In Table 8, the standardized coefficients for each variable are listed. These new beta

coefficients are helpful for comparing multiple regressors to each other without the confusion of

different units and measurements of the previous models. Instead, each of these coefficients

deals in standard deviations. If the X variable increases by one standard deviation, the Y

variable will increase by the amount of the coefficient (Gujarati 158). A larger coefficient in

comparison to another means that the former contributes more to the explanation of the Y

variable than the latter (Gujarati 158).

It is interesting to compare the relative strength of each variable to the others, especially

considering the wide breadth of measurements (from percentages to dollars) and units (from

decimals to billions) that the variables take on in their unstandardized forms.

When comparing these standardized variables it appears that female professors have the

greatest relative strength. Given the extremely low p-value for FemProf of 0.00126, it makes

sense that it would be a strong regressor when compared to the other variables with slightly

higher p-values. Female unemployment rates have the lowest relative strength. This is not

surprising considering female unemployment rates were found to be statistically insignificant in

Regression One.

Standardized coefficients were found by translating the data into standardized variables using the equation:

*=(Y

- 



)/S

and X

*=(X

- 



)/S

. These new data points were regressed to give the standardized betas above.

Possible OLS Assumption Errors

According to the table in Appendix B, there appears to be some multicollinearity amongst

a number of the explanatory variables in the model. Multicollinearity arises when there are

linear relationships among the X variables (Gujarati 321). In order to define two variables as

being highly correlated, the absolute value of their correlation coefficients must be above .7.

Multicollinearity is undesirable in a model because it is difficult to make precise

estimations of the betas when it is present (Gujarati 327). This can lead to the variables

presenting as statistically insignificant and also creating an artificially high goodness of fit

measurement (Gujarati 327).

By examining the table in Appendix B, it is clear that there are a number of relationships

which show multicollinearity. For example, in the correlation coefficients for Regression One,

the variable FemProf is highly correlated with FedFin with a correlation coefficient of .979 and

SATMath with a correlation coefficient of .889. Additionally, FedFin is highly correlated with

SATMath with a correlation coefficient of .900 and WEEA is highly correlated with SATMath

with a correlation coefficient of -.708.

There are a number of reasons that multicollinearity may appear. It may be that there is a

model specification error. This means that the linear model chosen to estimate the dependent

variable in this paper is not the best model to choose for the data set. However, because the data

used in this paper is time series data, it is very likely that multicollinearity is present simply due

to the fact that many of the variables follow similar trends. A number of the variables chosen for

this model have trended upward and thus will appear to be related to one another because they

are moving in the same direction. This may not mean that they have any bearing on each other;

it may simply be that because the data is moving in similar directions, they only appear to be

related.

Forecasting

One reason for developing an economic model is to use it for forecasting and prediction.

The resulting forecast can aid in developing policy and making decisions about the future.

It is important, however, to keep in mind that forecast error can occur for many reasons

when using an estimated model for forecasting. There may have been an error that occurred

when developing the model, leading any predictions formulated by that model to be skewed.

There may also be errors in the values used to represent the X independent variables. It is

difficult to accurately predict what will happen in the future, therefore predicting the exact

correct values for every independent variable in a model can be unlikely.

Another potential problem to consider when using an estimated model for forecasting is

that the potential for forecast error grows the further in the future you try to predict. Because a

model, and especially in this case of a single-equation regression model, is developed using

specific points, using the same model for points outside of the calculated region will naturally

produce some level of forecast error.

With all of this in mind, I will attempt to use the model found in Regression One stated in

the functional form:

FemSTEM = ƒ (FemProf, FedFin, SATMath, WEEA, FemUnemp)

to predict the percentage of females in STEM majors for the years 2011 and 2012. To do this I

found the known and forecasted values for each of the independent variables for the years 2011

and 2012. These values are displayed in Table 9.

Table 9: Forecasted Values for Regression One

FemProf

FedFin

SATMath

WEEA

FemUnemp

2011 110.5* $33.013 500 $2.521 8.5

2012 116* $32.188*

499 $2.470 7.9

in thousands

in billions, converted to 2014 dollars

in millions, converted to 2014 dollars

*these values were speculated using research, all other values are known

The predicted FedFin value for 2012 was formulated through a number of assumptions.

The final funds from the American Recovery and Reinvestment Act were distributed in 2010.

Due to the absence of ARRA funding, a large decrease in funding of about 11 percent from 2010

to 2011 is visible (“Federal Science and Engineering Obligations…”). Prior to this abnormally

large drop in funding, federal obligations towards science and engineering support had been

decreasing at a rate of about 2 to 3 percent every year from 2006 to 2010, not including the large

spike in 2009 explained by the ARRA stimulus. Therefore, the value speculated for 2012

reflects a 2.5 percent decrease in funding from 2011 to follow the general decreasing trend of the

previous years.

The number of female professors has been increasing by varying degrees every year since

1973. The values for FemProf for 2011 and 2012 were calculated by following the trends from

2009 and 2010 of approximately a 5 percent increase.

The values from Table 9 were plugged into the equation using the coefficients from

Regression One

. The forecast for the percent of females who choose to major in STEM fields in

2011 is 29.494 percent. The forecast for the percent of females who choose to major in STEM

fields in 2012 is 29.909 percent. Comparing these numbers to the previously known values used

FemSTEM = -.0276 + .000658FemProf – 0.00160FedFin + 0.00111SATMath + 0.000369WEEA –

0.000574FemUnemp

for the regression equations, the percent of females who will choose to major in STEM fields is

predicted to increase.

In 2011 there is predicted to be a 1.82 percent increase in female STEM majors from

27.677 percent in 2010 to the predicted 29.494 percent in 2011. This is due in part to the

predicted increase of female faculty members in both 2011 and 2012. It may also be due to the

fact that federal funding is expected to continue decreasing and because the relationship found

between FedFin and FemSTEM is negative, the double negative causes a positive increase in

FemSTEM.

In 2012 there is predicted to be a .41 percent increase in female STEM majors from the

forecasted 29.494 percent in 2011to 29.909 percent in 2012. The small change is due to the fact

that SAT math scores decrease and offset some of the increase gained by the predicted rising

number of female faculty members as well as the further decrease predicted for federal funding.

A decrease in WEEA appropriations also contributes to the low level of increase in female

STEM majors from 2011 to 2012.

Summary

Based on the two regressions, it is can be seen that five of the seven variables tested were

statistically significant

when regressed with the percentage of females in STEM majors.

Average salary and female unemployment rates were found to be statistically insignificant to the

explanation of the dependent variable. Additionally federal financial obligations, though

statistically significant, had a negative relationship with the dependent variable which was

opposite of the predicted model.

This model may be useful for policy in the future when legislators are looking for ways to

bolster female participation in STEM fields as a way to boost gender equity and the economy.

Those variables which were statistically significant were: FemProf,FedFin, NAEP, SATMath, and WEEA.

As seen in the section on forecasting, female STEM majors are expected to have increased in

2011 and 2012.

It will be important for legislators to note that federal funding must be timed correctly

and administered to the appropriate sectors in order to make the desired impact. This implication

can be seen in the negative relationship found between the FedFin variable and FemSTEM. The

negative relationship between these two variables indicates that more federal funding towards

STEM fields in universities may actually decrease the number of females choosing STEM

majors. Though this outcome may be due to errors in the regression model, it is still an

interesting and worrisome outcome which should not be taken lightly.

Additionally, policymakers should note that early concentration in mathematics is a

statistically significant indicator of female STEM majors. By building a foundation of

quantitatively-literate female students, the United States may then be able to increase the number

of females in STEM majors and thus in STEM occupations.

If given more time and resources, I would like to improve this study through a number of

methods. First, I would like to find a different, more consistent set of data for the AvgSal

variable. The data available to me for this variable was limited and therefore required that I pull

data points from a number of different reports. This could mean that the different data points

were not measured or collected in the same way and could have resulted in the outcome of

statistical insignificance. With a consistent set of data from one report, there may be a different

regression outcome.

Second, I would like to attempt to rerun Regression One, offsetting the variables of

FedFin and FemUnemp in order to try to correct for the unpredicted sign and statistical

insignificance, respectively. Perhaps by correlating one year’s data to the FemSTEM data of a

number of years later, I would receive the expected results.

Lastly, I would like to explore the effect that cultural trends and attitudes have on the

number of females choosing STEM majors. Historically, there have been cultural pressures for

females to focus on softer subjects such as the humanities and for males to focus on the hard

sciences such as chemistry and mathematics. The lasting effect that these pressures have on

females’ choice of major is difficult to quantify, especially with macro data like the sets that I

was using for this study. Many of the studies that I came across which discussed the social and

cultural factors in choice of major were pursued with data collected from personal surveys rather

than aggregated, impersonal data. If given the time and resources, I would like to attempt to

capture and quantify such variables as public attitude towards genders in specific fields, parental

support for females in STEM subjects, and personal perception of ability to achieve in STEM

subjects among women.

Appendix A

Percentage of Female Stem Majors

Source: National Science Foundation/National Center for Science and Engineering Statistics;

data from Department of Education/National Center for Education Statistics: Integrated

Postsecondary Education Data System Completions Survey

Year FemSTEM

1973 23.486%

1974 24.223%

1975 24.292%

1976 24.401%

1977 24.468%

1978 24.518%

1979 24.428%

1980 24.536%

1981 24.509%

1982 24.954%

1983 24.773%

1984 25.047%

1985 25.559%

1986 25.617%

1987 25.369%

1988 24.949%

1989 24.600%

1990 24.725%

1991 24.764%

1992 25.531%

1993 25.823%

1994 26.464%

1995 27.349%

1996 27.820%

1997 28.279%

1998 28.254%

1999 28.110%

2000 27.966%

2001 28.035%

2002 28.050%

2003 28.337%

2004 28.290%

2005 28.209%

2006 27.798%

2007 27.538%

2008 27.514%

2009 27.457%

2010 27.677%

Mean 26.150%

StdDev 1.5967%

Calculated by dividing females receiving bachelor degree in STEM fields by total number of females receiving

bachelor degrees.

Female Professors

Source: National Science Foundation, National Center for Science and Engineering Statistics,

special tabulations (2013) of the Survey of Doctorate Recipients (various years).

Year FemProf

1973 10.7

1974 12.2

1975 13.6

1976 15.1

1977 16.5

1978 18.0

1979 19.4

1980 21.3

1981 23.1

1982 24.8

1983 26.5

1984 28.8

1985 31.1

1986 32.6

1987 34.0

1988 36.4

1989 38.7

1990 40.3

1991 41.9

1992 44.4

1993 46.9

1994 49.7

1995 52.4

1996 55.8

1997 59.2

1998 61.8

1999 64.4

2000 67.5

2001 70.5

2002 74.6

2003 78.7

2004 84.4

2005 87.3

2006 90.1

2007 92.2

2008 94.2

2009 99.7

2010 105.2

Mean 49.0

StdDev 27.7414

In thousands, every other year beginning 1974 is an average

Federal Financial Obligations

Source: National Science Foundation/National Center for Science and Engineering Statistics,

Survey of Federal Science and Engineering Support to Universities, Colleges, and Nonprofit

Institutions.

Year FedFin

1973 $13.031

1974 $13.034

1975 $12.244

1976 $12.212

1977 $12.983

1978 $14.268

1979 $14.446

1980 $13.650

1981 $13.076

1982 $12.601

1983 $13.390

1984 $14.254

1985 $15.836

1986 $15.919

1987 $17.708

1988 $18.135

1989 $19.086

1990 $18.809

1991 $20.526

1992 $21.483

1993 $20.852

1994 $21.968

1995 $22.277

1996 $21.622

1997 $22.082

1998 $23.179

1999 $25.446

2000 $27.101

2001 $29.833

2002 $31.883

2003 $34.017

2004 $33.978

2005 $33.710

2006 $32.916

2007 $31.911

2008 $30.996

2009 $39.578

2010 $38.037

Mean $21.792

StdDev $8.174

In billions, converted to 2014 dollars

Average NAEP Math Scores for Fourth Grade Females

Source: U.S. Department of Education, National Center for Education Statistics, National

Assessment of Educational Progress (NAEP), NAEP 2012 Trends in Academic Progress; and

2012 NAEP Long-Term Trend Mathematics Assessment, retrieved August 29, 2013, from Long-

Term Trend NAEP Data Explorer.

Year taken Score

1973

220

1978

220

1982

221

1986

222

1990

230

1992

228

1994

230

1996

229

1999

231

Mean

226

StdDev

Average SAT Math Scores for Females

Source: College Board: 2013 College-Bound Seniors Total Group Profile Report

Year SATMath

1973 489

1974 488

1975 479

1976 475

1977 474

1978 474

1979 473

1980 473

1981 473

1982 473

1983 474

1984 478

1985 480

1986 479

1987 481

1988 483

1989 482

1990 483

1991 482

1992 484

1993 484

1994 487

1995 490

1996 492

1997 494

1998 496

1999 495

2000 498

2001 498

2002 500

2003 503

2004 501

2005 504

2006 502

2007 499

2008 499

2009 498

2010 499

Mean 487.263

StdDev 10.246

Median Salary for STEM Occupations

Source: National Science Foundation, Division of Science Resources Statistics, Scientists and

Engineers Statistical Data System (SESTAT),

http://sestat.nsf.gov.

Year

Salary

1993

$77,989

1994

$77,626

1995

$77,027

1996

$78,559

1997

$80,454

1998

$82,821

1999

$84,554

2000

$84,259

2001

$84,433

2002

$85,565

2003

$86,134

2004

$84,491

2005

$84,173

2006

$83,990

2007

$81,528

2008

$81,730

2009

$83,828

2010

$83,983

Mean

$82,397

StdDev

$2,806.95

In 2014 dollars

Appropriations to Women’s Educational Equity Act

Source: U.S. Department of Education

Year WEEA

1973 $0.000

1974 $0.000

1975 $0.000

1976 $25.871

1977 $28.165

1978 $29.113

1979 $29.105

1980 $28.492

1981 $20.985

1982 $14.013

1983 $13.577

1984 $13.015

1985 $13.091

1986 $12.296

1987 $7.233

1988 $6.650

1989 $5.583

1990 $3.768

1991 $3.438

1992 $0.836

1993 $3.223

1994 $3.143

1995 $3.178

1996 $0.000

1997 $4.388

1998 $4.321

1999 $4.227

2000 $4.090

2001 $3.979

2002 $3.915

2003 $3.802

2004 $3.681

2005 $3.553

2006 $3.408

2007 $2.128

2008 $2.013

2009 $2.615

2010 $2.608

Mean $8.250

StdDev $8.972

In millions, in 2014 dollars

Female Unemployment Rates

Source: 2013 Economic Report of the President

Year FemUnemp

1973 6.0

1974 6.7

1975 9.3

1976 8.6

1977 8.2

1978 7.2

1979 6.8

1980 7.4

1981 7.9

1982 9.4

1983 9.2

1984 7.6

1985 7.4

1986 7.1

1987 6.2

1988 5.6

1989 5.4

1990 5.5

1991 6.4

1992 7.0

1993 6.6

1994 6.0

1995 5.6

1996 5.4

1997 5.0

1998 4.6

1999 4.3

2000 4.1

2001 4.7

2002 5.6

2003 5.7

2004 5.4

2005 5.1

2006 4.6

2007 4.5

2008 5.4

2009 8.1

2010 8.6

Mean 6.4

StdDev

1.473

Appendix B

Simple Correlation Coefficients among Variables

Regression One: Correlation coefficients between variables

FemSTEM

FemProf FedFin SATMath

WEEA FemUnemp

FemSTEM 1

FemProf 0.885182

FedFin 0.853025

0.978654

SATMath 0.875535

0.88666

0.900483

WEEA -0.50119

-0.54425

-0.53971

-0.70815

FemUnemp

-0.60816

-0.49743

-0.47854

-0.67068

0.474385

Regression Two: Correlation coefficients between variables

NAEP STEM

NAEP 1

STEM 0.735345077

Correlation coefficients including AvgSal

FemSTEM

FemProf FedFin SATMath WEEA FemUnemp

Salary

FemSTEM 1

FemProf 0.334491

FedFin 0.337673

0.937851

SATMath 0.759997

0.783752

0.793543

WEEA 0.29071

-0.15304

-0.02275

0.165071

FemUnemp

-0.39378

0.329634

0.405831

-0.14596

-0.30304

Salary 0.690664

0.621215

0.723629

0.856353

0.426234

-0.08967

Appendix C

Final Regression Results

SUMMARY OUTPUT: Regression One

Regression Statistics

Multiple R 0.929438546

R Square 0.863856012

Adjusted R Square 0.842583513

Standard Error 0.006420133

Observations 38

ANOVA

df SS MS F

Significance

Regression 5

0.008369142

0.001673828

40.60905323

6.11075E-13

Residual 32

0.001318979

4.12181E-05

Total 37

0.009688121

2.021

Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%

Intercept

0.275894178

0.17688016

1.559780235

0.128648275

0.636187274

0.084398918

0.636187274

0.084398918

FemProf 0.000658454

0.000186217

3.535949866

0.001263473

0.000279142

0.001037766

0.000279142

0.001037766

FedFin

0.001603741

0.000718547

2.231922864

0.032747961

0.003067373

0.000140109

0.003067373

0.000140109

SATMath 0.001109658

0.000366833

3.024968422

0.004873387

0.000362444

0.001856872

0.000362444

0.001856872

WEEA 0.000369661

0.0001761

2.099159358

0.043779685

1.0958E-05

0.000728365

1.0958E-05

0.000728365

FemUnemp

0.000573986

0.001061248

-0.54085961

0.592349722

0.002735679

0.001587706

0.002735679

0.001587706

SUMMARY OUTPUT: Regression Two

Regression Statistics

Multiple R 0.735345077

R Square 0.540732382

Adjusted R Square

0.464187779

Standard Error 0.009869705

Observations 8

ANOVA

df SS MS F Significance F

Regression 1

0.000688139

7.064278362

0.037630853

Residual 6

0.000584466

9.74111E-05

Total 7

0.001272605

Coefficients Standard Error

t Stat P-value Lower 95% Upper 95% Lower 95.0%

Upper 95.0%

Intercept -0.2111025

0.181351628

-1.164050759

0.28859314

-0.654853947

0.232648947

-0.654853947

0.232648947

NAEP 0.002141867

0.000805858

2.657871021

0.037630853

0.000170003

0.00411373

0.000170003

0.00411373

SUMMARY OUTPUT: Regression Including Average Salary

Regression Statistics

Multiple R 0.910342625

R Square 0.828723696

Adjusted R Square 0.735300257

Standard Error 0.003403204

Observations 18

ANOVA

df SS MS F Significance F

Regression 6

0.000616426

0.000102738

8.870618621

0.001087901

Residual 11

0.0001274

1.15818E-05

Total 17

0.000743826

2.11

Coefficients Standard Error

t Stat P-value Lower 95% Upper 95% Lower 95.0%

Upper 95.0%

Intercept -0.72723457

0.234573224

-3.100245448

0.010099715

-1.243526756

-0.210942385

-1.243526756

-0.210942385

Faculty -3.61839E-05

0.000159363

-0.227053482

0.82454673

-0.000386939

0.000314572

-0.000386939

0.000314572

FedOb -0.001371798

0.000638484

-2.148522865

0.054787512

-0.002777093

3.34964E-05

-0.002777093

3.34964E-05

SAT Math 0.001962511

0.000527865

3.717828363

0.003394582

0.000800688

0.003124333

0.000800688

0.003124333

WEEA -0.000221824

0.001055672

-0.210126124

0.837410633

-0.002545342

0.002101693

-0.002545342

0.002101693

FemUnemp 0.002178751

0.001397604

1.558918392

0.147304125

-0.000897355

0.005254858

-0.000897355

0.005254858

Salary 7.39735E-07

7.51269E-07

0.984647756

0.345959614

-9.13796E-07

2.39327E-06

-9.13796E-07

2.39327E-06

Works Cited

Amy Bergerson, et al. "Gender Differences In Expressed Interests In Engineering-Related Fields

ACT 30-Year Data Analysis Identified Trends And Suggested Avenues To Reverse

Trends." Journal of Career Assessment 21.4 (2013): 599-613. Academic Search

Complete. Web. 20 Oct. 2013.

Arcidiacono, Peter, Joseph V. Hotz, and Songman Kang. "Modeling College Major Choices

Using Elicited Measures of Expectations and Counterfactuals." Journal of Econometrics

166.1 (2012): 3-16. Web. 31 Mar. 2014.

Beede, David, Tiffany Julian, David Langdon, George McKittrick, Beethika Khan, and Mark

Doms. Women in STEM: A Gender Gap to Innovation. Rep. N.p.: n.p., n.d. U.S.

Department of Commerce. Web.

Carnevale, Anthony P., Ban Cheah, and Jeff Strohl. Hard Times: College Majors,

Unemployment and Earnings: Not All College Degrees Are Created Equal. Rep. Center

on Education and the Workforce, 4 Jan. 2012. Web. 26 Mar. 2014.

Calkins, Lindsay. Welki, Andrew. “The Economics Major: How Students Decide.” March 2003.

Web. 13 Oct. 2013.

"Federal Science and Engineering Obligations to Universities and Colleges Drop by 11% in FY

2011." National Science Foundation, National Center for Science and Engineering

Statistics. N.p., Mar. 2014. Web. 23 Apr. 2014.

"Fiscal Year 2012 Budget Summary." U.S. Department of Education. N.p., 14 Feb. 2011. Web.

22 Apr. 2014.

Griffith, Amanda L. "Persistence Of Women And Minorities In STEM Field Majors: Is It The

School That Matters?." Economics Of Education Review 29.6 (2010): 911-922. Academic

Search Complete. Web. 5 Dec. 2013.

Gujarati, Damodar N., and Dawn C. Porter. Basic Econometrics. Boston: McGraw-Hill Irwin,

2009. Print.

Heston-Demirel, Vivienne. "Women's Educational Equity Act (WEEA)." National Counsel for

Research on Women, 15 July 2010. Web. 25 Mar. 2014.

Mazza, Jacoppo. "Degree Choices." Royal Economic Society. N.p., n.d. Web. 31 Mar. 2014.

S. 3475, 112 Cong. (2012) (enacted). Print.

"Subpart 21 - Women's Educational Equity Act." U.S. Department of Education. N.p., 15 Sept.

2004. Web. 25 Mar. 2014.

Stout, Jane, Nilanjana Dasgupta, Matthew Hunsinger, and Melissa McManus. "STEMing the

Tide: Using Ingroup Experts to Inocoluate Women's Self-Concept in Science,

Technology, Engineering, and Mathematics (STEM)." Journal of Personality and Social

Psychology 100.2 (2011): 255-70. Web.

"Title IX of the Education Amendments of 1972." Title IX of the Education Amendments of

1972. N.p., n.d. Web. 01 Dec. 2013.

"Title IX and Sex Discrimination." US Department of Education. N.p., 18 June 2012. Web. 24

Mar. 2014.

U.S. Department of Education. Women’s Educational Equity. CFDA No. 84.083. Biennial

Evaluation Report – FY 93-94. <https://www2.ed.gov/pubs/Biennial/125.html>

Accessed: 3/25/2014

"Women in STEM." The White House. N.p., n.d. Web. 03 Dec. 2013.

Young, Danielle M., Laurie A. Rudman, Helen M. Buettner, and Megan C. McLean. "The

Influence of Female Role Models on Women's Implicit Science Cognitions." Psychology

of Women Quarterly (2013): 283-92. SAGE Publications, 11 Apr. 2013. Web. 06 Dec.

2013.